Date: Thu, 21 Sep 2000 22:18:24 -0400 (EDT)
From: James Wadsley <wadsley AT physun DOT physics DOT mcmaster DOT ca>
To: pgcc AT delorie DOT com
Subject: pgcc didn't perform for me
Message-ID: <Pine.GSO.4.21.0009212159160.24859-100000@physun.physics.mcmaster.ca>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: pgcc AT delorie DOT com


I am not on the mailing list.

I have a high performance tree-code for astrophysical problems and I was
running on an AMD-K7 900 MHz.

I tried first gcc 2.95.2 and then the pgcc version using the patch to
convert. 

Using simply -O3 gcc 2.95.2 was about 10% faster than pgcc with -O3.
Using the suggested optimizations for pgcc (align double, funroll-loops
and -mk6) the code simply crashed with a segmentation fault.

I was quite disappointed -- can you suggest a problem in what I attempted?

You should put some standard benchmarks on your site.
After my tests I'm not going to simply take your word for
it that there is an advantage to pgcc.

Sorry,

James Wadsley.